| Abstract: | 幾十年來許多國家面臨了有關人口老化的挑戰，其中老人晚年的憂鬱症狀對身體健康及功能的嚴重影響已經引起廣泛的關注。有鑒於此，本研究利用行政院衛生署國民健康局所提供之1989-2007 年六波「台灣地區老人保健與生活問題長期追蹤調查」資料針對影響老人憂鬱狀態變化趨勢的相關因素，包括人口特徵、家庭與環境條件及健康狀況進行深入的探討研究。研究方法上，基於憂鬱狀態復發具有區間設限的特性，在同一個體於不同區間之事件復發視為相互獨立的假設下，利用區間設限模式探討影響老人憂鬱變化趨勢相關因子，惟區間設限模式的缺點在於無法處理兩種以上事件同時存在的情形，因此，本研究利用具馬可夫鏈特性之Cox 及Aalen 多重狀態事件史模式，探討有關老人憂鬱狀態變化趨勢。有關Cox 比例風險模式假設之驗證，除利用Aalen 加成模式進行輔助探討時間相依變數的特性外，亦藉由Schoenfeld 殘差圖進行交叉驗證。另一方面，老人憂鬱狀態變化趨勢之轉移機率亦經由Cox 及Aalen 多重狀態模式配適結果進行估算。本研究結果發現年齡、性別、教育程度、有無配偶、與子女同住、經濟狀況、健康自評、日常生活活動能力（ADL）及體能狀況與老人憂鬱狀態變化趨勢有密切的相關；在轉移機率方面，發現初始狀態已有憂鬱之老人，其死亡率相較於無憂鬱之老人有偏高的趨勢。
Many countries have been facing the challenges of aging population over decades. Depressive symptoms in later life of elderly have serious implications for the health and functioning has caused wide public concern. The purpose of this study is to explore the effect of the covariates related to demographic characteristics, home and environment conditions, and health status, on the changing status of depression of old people in Taiwan. A representative panel sample survey data collected in the six waves of “The Longitudinal Sample Survey of Health and Living Status of the Elderly in Taiwan” conducted from 1989 to 2007 by the predecessor organization of the Bureau of Health Promotion, Department of Health is used for analysis. Since cases of depression status are recurrent events with the characteristics of interval censored data, Cox model with interval-censored approach is applied for investigating the variables related to the changing status of depression among the elderly under the assumption that the observations within a subject are mutually independent. However, a major disadvantage of interval censored model is unable to deal with the situation that two or more different events of interest exist simultaneously, a more general method, multi-state model considered as a Markov chain for event history analysis is used to interpret the changing status of depression. To verify the proportional assumption in the conventional Cox model, Aalen’s additive model taking into account instantaneous covariate effects in time is employed as an alternative on a supplement. Also, the scaled Schoenfeld residuals plots are used for investigating the proportional hazard assumption. Furthermore, the transition probabilities for the changing status of depression obtained from Cox and Aalen models are provided. From the results of this study, age, gender, education, spouse, living with children, economic status, self-rated health, ADL function and physical function are significantly related to the depression. Also, the transition probabilities show the death rate has a greater impact on the elderly with initial depression state. |
